#2010-005

Is there complementarity or substitutability between internal and external R&D strategies?

John Hagedoorn and Ning Wang

Working Paper Series

United Nations University - Maastricht Economic and social Research and training centre on Innovation and Technology

Keizer Karelplein 19, 6211 TC Maastricht, The Netherlands Tel: (31) (43) 388 4400, Fax: (31) (43) 388 4499, e-mail: info@merit.unu.edu, URL: http://www.merit.unu.edu

UNU-MERIT Working Papers ISSN 1871-9872

Maastricht Economic and social Research and training centre on Innovation and Technology, UNU-MERIT

UNU-MERIT Working Papers intend to disseminate preliminary results of research carried out at the Centre to stimulate discussion on the issues raised.

IS THERE COMPLEMENTARITY OR SUBSTITUTABILITY BETWEEN

INTERNAL AND EXTERNAL R&D STRATEGIES?

John Hagedoorn

Department of Organization and Strategy and UNU-MERIT

Maastricht University

The Netherlands

Tel (31) 43-3883823

Fax (31) 43-3884893

J.Hagedoorn@maastrichtuniversity.nl

Ning Wang

Department of Organization and Strategy

Maastricht University

The Netherlands

Tel (31) 43-3884793

Fax (31) 43-3884893

N.Wang@maastrichtuniversity.nl

January 8, 2010

JEL codes: O32, L24

2

Is There Complementarity or Substitutability between Internal and External

R&D Strategies?

Abstract

The mixed picture of extant research on the relationship between internal and external R&D

prompts us to ask such a question: under what conditions is there complementarity or

substitutability between different R&D strategies? The goal of this paper is to contribute to the

empirical literature by advancing and testing the contingency of the relationship between internal

and external R&D strategies in shaping firms‘ innovative output. Using a panel sample of

incumbent pharmaceutical firms covering the period 1986−2000, our empirical analysis suggests

that the level of in-house R&D investments, which is characterized by decreasing marginal

returns, is a contingency variable that critically influences the nature of the link between internal

and external R&D strategies. In particular, internal R&D and external R&D, through either R&D

alliances or R&D acquisitions, turn out to be complementary innovation activities at higher levels

of in-house R&D investments, whereas at lower levels of in-house R&D efforts internal and

external R&D are substitutive strategic options. These findings are robust to alternative

specifications and estimation techniques, including a dynamic perspective on firm innovative

performance.

Key words: Complementarity; Substitutability; Internal R&D; External R&D; Innovative output;

Pharmaceutical Industry; Biotechnology

3

INTRODUCTION

The ability of firms to continually update their technological know-how and capabilities is

becoming an imperative for competitive survival (Christensen, 2000; Foster and Kaplan, 2001).

In that context, firms can pursue multiple approaches to innovative renewal. One path is to

develop and nurture in-house research and development (R&D) competencies. In technology-

based industries, perhaps the most salient mechanism of firms‘ innovative renewal is investing in

internal R&D (Knott and Posen, 2009). Another path is found in the use of external technology

sourcing. Powell et al. (1998, 1996) point out that in industries characterized by complex and

rapidly expanding knowledge bases, the locus of innovative renewal lies within a broad

‗network‘ of learning rather than within the boundaries of individual firms (see also Breschi and

Malerba, 1997; Edquist, 1997). As a consequence, incumbent firms frequently need to leverage

their external networks to source new technology and capabilities, especially within an emerging,

new technological regime (Bettis and Hitt, 1995; Kranenburg and Hagedoorn, 2008; Nicholls-

Nixon and Woo, 2003; Rothaermel and Hess, 2007).

Recent years have witnessed more and more companies pursuing an ‗open innovation‘

(Chesbrough, 2003) strategy by leveraging internal and external knowledge flows in parallel to

build and hone their innovative capabilities. Concurrent with the open innovation approach, a

rather voluminous literature has emerged that examines the relationship between in-house R&D

and external technology sourcing. At the heart of this literature is the discussion of the

complementarity or substitutability between internal and external R&D strategies for managing

innovation, a debate that has been accompanied by mixed empirical evidence1. On the one hand,

1 The study of complementarity between activities can be traced back to the theory of supermodularity (Milgrom and

Roberts, 1990, 1995). Complementarity implies that adding one activity while the other is already being performed

will result in higher incremental innovative performance than when adding the activity in isolation, whereas

4

a diverse set of studies demonstrate that internal R&D and external technology sourcing are

complementary innovation activities, suggesting the interrelatedness of internal and external

R&D in improving innovative performance (Brown and Eisenhardt, 1997; Caloghirou et al., 2004;

Cassiman and Veugelers, 2006; Lokshin et al., 2008; Schmiedeberg, 2008; Tripsas, 1997; Tsai

and Wang, 2008; Veugelers, 1997). And on the other, a host of empirical work identifying the

linkage between internal and external sources of innovation have found substitutability instead

(Audretsch et al., 1996; Blonigen and Taylor, 2000; Higgins and Rodriguez, 2006; Laursen and

Salter, 2006; Watkins and Paff, 2009). The above mixed picture of extant research on the

relationship between internal and external R&D prompts us to ask the question: under what

conditions is there complementarity or substitutability between different R&D strategies? The

goal of this paper is to contribute to the empirical literature by advancing and testing the

contingency of the relationship between internal and external R&D strategies in shaping firms‘

innovative output.

Our study focuses on the global pharmaceutical industry as it provides an ideal setting for

us to explore how incumbent firms organize R&D strategies in their quest for innovation within a

new biotechnology-based technological regime. Using a panel sample of 83 incumbent

pharmaceutical firms covering the period 1986–2000, our empirical analysis suggests that the

level of in-house R&D investments, which is characterized by decreasing marginal returns, is a

contingency variable that critically influences the nature of the link between internal and external

R&D strategies. In particular, internal R&D and external R&D, through either R&D alliances or

R&D acquisitions, turn out to be complementary innovation activities at higher levels of in-house

substitutability refers to the case in which adding one activity can actually decrease the marginal or incremental

innovative performance of other activities (Cassiman and Veugelers, 2006; Mohnen and Röller, 2005).

5

R&D investments, whereas at lower levels of in-house R&D efforts internal and external R&D

are substitutive strategic options.

The paper proceeds as follows. After an overview of the growing literature on internal and

external R&D strategies of firms, we formulate the hypotheses to be tested. We then turn to

methodology, describing the sample, the model, and the estimation techniques. The empirical

results are presented in the next section, followed by a variety of robustness tests. The paper

concludes with a discussion of the results and some implications for future research.

THEORY AND HYPOTHESES

To the extent that R&D is an important, and perhaps even the most important, contributor to

productivity growth and innovation (Griliches, 1979; Scherer, 1982), the impact of R&D

investments on firm innovation has attracted enormous attention in the literature. In their seminal

work on the R&D-patents relationship, Hausman et al. (1984) maintain that, rather than the

propensity to patent just declining exogenously over time, firms are getting less patents from their

more recent R&D investments, implying a decline in the ‗effectiveness‘or productivity of R&D.

The role of R&D in patenting, as a major indicator of innovation, has later on been studied

extensively (Blundell et al., 1995; Blundell et al., 2002; Crépon and Duguet, 1997; Guo and

Trivedi, 2002; Montalvo, 1997), with the overall R&D elasticity of patents varying from 0.4 to

0.7 and thereby suggesting decreasing returns to scale (Gurmu and Perez-Sebastian, 2007).

Amongst others, Griliches (1990) suggests that the estimates for R&D elasticity vary and

depend on firm size, with larger firms showing a lower R&D productivity2 (see also Acs and

Audretsch, 1991). In line with this thinking, Graves and Langowitz (1993) show that firms in the

2 Griliches (1990) also points to the possibility of a reverse causality from R&D to patents in the knowledge

production function. Consequently, R&D is unlikely to be strictly exogenous.

6

pharmaceutical industry experience decreasing returns to scale in R&D as the level of their R&D

expenditures rises. Their central argument is that innovative effectiveness decreases with

increasing firm R&D investments and, by association, with firm size. Cohen and Klepper (1996)

explain why larger firms can still prosper despite the lower average productivity of their R&D, as

the greater output over which large firms can apply their R&D enables them to profit more from

R&D than smaller firms, which leads them to undertake more R&D projects at the margin. A

large number of recent studies examine the impact of firm R&D investments on innovation as

well (e.g., Griffith et al., 2004; Hagedoorn and Duysters, 2002; Lokshin et al., 2008; Love and

Roper, 2002; Rothaermel and Hess, 2007) and find diseconomies of scale in R&D in most cases.

The above leads us to a baseline hypothesis on the effect of internal R&D on innovative output,

which is to be seen as a starting point for our analysis of the simultaneous interactions between

different R&D strategies.

Hypothesis 1: There are decreasing returns to scale in internal R&D. Firm innovative output

initially increases with investment in internal R&D, this rate of increase diminishes at higher

levels of internal R&D investments though.

Openness of firms to external knowledge sources is another key element for knowledge

creation and innovation (Arora and Gambardella, 1990; Cohen and Levinthal, 1990). A diverse

set of studies emphasize that access to external R&D can be leveraged to enhance the efficiency

of internal R&D investments. Brown and Eisenhardt (1997) demonstrate that firms that leverage

external technology sourcing as a complement to internal R&D investments in order to probe and

access cutting-edge knowledge are most successful in their new product introductions. Tripsas

(1997) finds, while studying the continued survival of incumbent firms when confronted with

7

radical technical change, that a combination of internal R&D and external technology sourcing

through alliances positively reinforces firms‘ innovative output3. In a similar fashion, Caloghirou

et al. (2004) show the parallel positive role of internal R&D capabilities and their interaction with

external sources of knowledge in raising innovative performance. Focusing on the choice

between make or buy decision by firms, Cassiman and Veugelers (2006) find that internal R&D

and external knowledge acquisition are complementary innovation activities. The moderating

effect of internal R&D is also emphasized by Tsai and Wang (2008). Their findings suggest that

the contribution of external technology acquisition to firm performance increases with the level

of internal R&D efforts4. In their dynamic panel study of Dutch manufacturing firms, Lokshin et

al. (2008) also find that combing internal and external R&D significantly contributes to

productivity growth, with a positive impact of external R&D only evident in case of sufficient

internal R&D.

However, a number of other papers identifying the relationship between in-house and

external sources of innovation have found a substitutability instead of a complementarity effect.

Audretsch et al. (1996) suggest that internal and external R&D are substitutes in low-technology

industries but not in high-technology industries. Blonigen and Taylor (2000) find a substantial

inverse relationship between R&D intensity and acquisition activities among electronic and

electrical equipment firms. Higgins and Rodriguez (2006) demonstrate that firms that are

experiencing deterioration in internal R&D productivity are more likely to engage in technology

acquisitions, implying that firms may pursue one mode of innovation to compensate for their

ineffectiveness in another mode. In their investigation into the role of firm openness in explaining

3 Also see the study by Schmiedeberg (2008), who provides further evidence for significant complementarities

between internal R&D and R&D cooperation, but casts doubts on the complementarity between internal and

contracted R&D. 4 This is consistent with the findings by Veugelers (1997) who shows that external knowledge sources stimulate the

innovative productivity of internal research activities at lease for firms with internal R&D departments.

8

innovative performance, Laursen and Salter (2006) report evidence of a substitution effect

between internal R&D and external search strategies. Consistent with previous studies,

mentioned in the above, though in a rather different research context, Watkins and Paff (2009)

find a substitute relationship between in-house R&D and external basic research when the

relative tax prices of each category of research change5.

By and large, these various strands of the empirical literature indicate the inconclusive

nature of the debate about the complementarity or substitutability between internal and external

R&D strategies in shaping firms‘ innovative performance. We therefore are faced with two

opposing hypotheses:

Hypothesis 2-a: Internal and external R&D strategies complement each other and hence

increase a firm’s innovative output.

Hypothesis 2-b: Internal and external R&D strategies substitute for each other and hence

decrease a firm’s innovative output.

METHODOLOGY

Research Setting

The research setting for testing our hypotheses is the global pharmaceutical industry, which is on

the verge of profound mutations as a new biotechnology-based technological regime has emerged.

More in particular, we will focus on incumbent pharmaceutical companies attempting to build up

innovative capabilities within this new technological regime. Incumbent pharmaceutical firms are

defined as pharmaceutical firms that were in existence prior to the emergence of biotechnology.

5 R&D policies such as R&D grants, subsidies, and tax incentives often treat different types of R&D differently,

which tends to affect the composition of firms‘ R&D budgets (Watkins and Paff, 2009). See similar studies by

Cappelen et al. (2008), Czarnitzki et al. (2004), Parsons and Phillips (2007).

9

These companies, such as Bayer, Hoffmann-La Roche, Merck, and Pfizer, are generally mature

and very large firms that dominated the industry since the 1940s. Since the mid 1970s, new

biotechnology brought along significant scientific and technological breakthroughs in genetic

engineering (recombinant DNA, 1973) and hybridization (monoclonal antibodies, 1975). These

advances have served as a radical process innovation for established pharmaceutical companies

in the way new drugs are discovered and developed (Pisano, 1997) and they are even to affect the

core capabilities needed to remain competitive (Nagarajan and Mitchell, 1998). This development

provides an ideal setting for us to explore how incumbents organize R&D strategies in their quest

for innovation within a new technological regime.

Data and Sample

We use data from several major sources. (1) The source for the patent data is the Technology

Profile Report maintained by the U.S. Patent and Trademark Office (USPTO), an agency of the

U.S. Department of Commerce. We obtained detailed data comprising the complete set of all

biotechnology patents granted to global pharmaceutical firms annually since the emergence of

biotechnology6. (2) The MERIT-CATI database, which is developed and maintained by

researchers at the Maastricht Economic Research Institute on Innovation and Technology

(MERIT). The CATI (Cooperative Agreements and Technology Indicators) database documents

6 The complete set of biotechnology patents refers to the biotechnology patents, as identified by the USPTO, in the

following patent classes: 424 [Drug, bio-affecting and body treating compositions (424/9.1-424/9.2, 424/9.34-

424/9.81, 424/85.1-424/94.67, 424/130.1-424/283.1, 424/520-424/583, 424/800-424/832) ], 435 [Chemistry:

Molecular biology and microbiology (435/1.1-435/7.95, 435/40.5-435/261, 435/317.1-435/975) ], 436 [Chemistry:

Analytical and immunological testing (436/500-436/829) ], 514 [Drug, bio-affecting and body treating compositions

(514/2-514/22, 514/44, 514/783) ], 530 [Chemistry: Natural resins or derivatives; peptides or proteins; lignins or

reaction products thereof (530/300-530/427, 530/800-530/868) ], 536 [Organic compounds (536/1.11-536/23.74,

536/25.1-536/25.2) ], 800 [Multi-cellular living organisms and unmodified parts thereof and related processes], 930

[Peptide or protein sequence], PLT [Plants].

10

technological partnerships7 in multiple industries around the globe. In this study we specifically

focused on R&D alliances between established pharmaceutical companies and new

biotechnology ventures. (3) The SDC (Securities Database Corporation) Platinum of Thomson

Financial is used for data on R&D acquisitions of new biotechnology companies by incumbent

pharmaceutical firms. This database contains information on the year of acquisition, details on

the (parent) acquirer and the (parent) target, and a brief description of the acquisition transaction.

(4) Financial data were cross-tracked through Compustat and Datastream (Thomson Financial).

All financial data were converted to U.S. dollars and inflation-adjusted.

To mitigate a potential survivor bias, we started with a comprehensive set of global

pharmaceutical firms alive in 1986 according to various industry sources8. In this manner we

identified 89 incumbent pharmaceutical firms, defined as pharmaceutical firms that were in

existence prior to the emergence of biotechnology in the mid 1970s. We then proceeded with a

family tree analysis9 on each of the 89 firms for the 1986–2000 time period. Through this

analytical process, 6 horizontal mergers and acquisitions were ascertained among the incumbent

pharmaceutical firms. As a horizontal merger or acquisition took place, we added to the survivor

the historical data of the acquired company (i.e. the inactive) and then tracked the surviving entity

forward10

. Such scrutiny procedure ultimately renders a slightly unbalanced panel sample of 83

firms covering the 15-year period from 1986 to 2000 (i.e., 1139 firm-year observations).

7 CATI only collects cooperative agreements in which a combined innovative activity or an exchange of technology

is engaged by at least two industrial partners. The first phase of data collection is described in Hagedoorn (1993),

Duysters and Hagedoorn (1993). 8 To draw the sample for this study we tracked Compustat, Datastream, Amadeus, SIC reports, Ernst and Young‘s

Annual Biotech Industry Reports, Scrip‘s Pharmaceutical Yearbook, amongst others. 9 For the rationale for constructing family trees, see Rothaermel and Hess (2007). When firms were acquired by

outsiders, they were included as unbalanced data. A variety of industry sources were checked throughout the family

tree analysis, including Compustat/Standard and Poor‘s, Encyclopedia, and company homepage. 10 In Compustat, the treatment of mergers and acquisitions seeks to preserve historical company data for the

acquiring company whenever possible. It is rare that a new company is added as a result of a merger. There are

usually no changes to the surviving company‘s company-level identifiers, whereas the acquired company

subsequently becomes inactive in the database the month following the merger or acquisition. Two basic accounting

11

It is important to indicate that while we collected patent data until the end of 2003, we

purposefully ended the study in 2000 to overcome a potential right truncation bias. This measure

was taken in that all patent data used in the study are based upon patent applications granted, and

the application date is only reported when the patent is actually granted. A grant lag as such is

generally three years after the patent was originally applied for11

(Biotechnology Industry

Organization, 2008). In contrast, there is essentially no lag time between the completed invention

and the patent application date, which is on average no more than three months (Darby and

Zucker, 2003). Thus, the application date of a granted patent is closely tied to the timing of the

new knowledge creation and should be used as the relevant time placer for patents (Hall et al.,

2001; Trajtenberg, 1990).

Measures

Dependent Variable

Innovative Output We focus on the patenting frequency, i.e., the number of annual

biotechnology patents (Biotech Patents) granted to incumbent pharmaceutical firms to proxy for

their innovative output within the new technological regime. Annual patent counts are generally

accepted as one of the most appropriate indicators that enable researchers to compare the

inventive or innovative output of companies in terms of new technologies, new processes, and

new products (Hagedoorn and Cloodt, 2003). Patents are directly related to inventiveness

methods are used to account for mergers and acquisitions, the ―purchase method‖ and the ―pooling of interest

method‖. In this study we principally followed the ―purchase method‖ since this is the way that the majority of

mergers and acquisitions are accounted for. In June 2001 the Financial Accounting Standard Board eliminated the

―pooling of interest‖ and requires that all business combinations be accounted for by the ―purchase method‖ (FASB

Statement 141). 11 The USPTO evaluates patent applications and grants patents. Patents usually last 20 years from the date on which

the patent application is filed (not when it is issued). Thus, the enforceable term of a patent is between 17 and 20

years; exactly how long depends on how long the application is under PTO review (i.e., the grant lag). The PTO

provides a three-year period for the agency to issue a patent.

12

(Walker, 1995), and represent an externally validated measure of technological novelty (Griliches,

1990). Studies also show that patents are well-correlated with other indicators of innovative

output (Hagedoorn and Cloodt, 2003; Stuart, 2000).

Independent Variables

Internal R&D We measure a pharmaceutical company‘s internal R&D investments by its

R&D expenditures. As highlighted in prior research, R&D expenditures are primarily taken as an

input indicator of a firm‘s efforts in establishing R&D that might eventually lead to innovative

output (Griliches, 1990, 1998; Hitt et al., 1997). R&D efforts can also signal innovative

competences that are found to affect the performance of companies, particularly in high-

technology industries (Duysters and Hagedoorn, 2001; Henderson and Cockburn, 1994). To

accurately identify the magnitude of in-house R&D investments by the global pharmaceuticals,

we excluded in-process R&D spending12

, which is subsumed within R&D expense in the

Compustat database. Meanwhile, the quadratic terms of internal R&D, R&D expenditures

squared, are enclosed in the regression analysis as a test for decreasing returns to scale in R&D13

(Blundell et al., 1995; Blundell et al., 2002; Cohen and Klepper, 1996; Crépon and Duguet, 1997;

Graves and Langowitz, 1993; Griffith et al., 2004; Griliches, 1990; Guo and Trivedi, 2002;

Hagedoorn and Duysters, 2002; Hall et al., 1986; Hausman et al., 1984; Lokshin et al., 2008;

Montalvo, 1997; Rothaermel and Hess, 2007).

12 Acquisition-related in-process R&D charges are defined in Compustat as the portion of R&D considered to be

‗purchased‘ and written off immediately upon acquisition if the R&D items are deemed not to have an alternative use. 13 Exclusion of the quadratic terms is nontrivial as in the presence of declining returns to scale in R&D one may

expect firms to avoid this by combing internal and external R&D strategies (Lokshin et al., 2008).

13

External R&D We include a dummy variable, rather than a count measure14

, indicating

whether or not a pharmaceutical firm pursued external R&D strategies. This dummy variable

takes on the value of 1 if the company sourced R&D externally through either R&D alliances or

R&D acquisitions (1 = R&D sourced externally through R&D alliances/R&D acquisitions), and

0 if the company solely undertook internal R&D. We identified R&D alliances through a detailed

content analysis of each and every alliance the pharmaceutical firms entered since 1986 in

biotechnology, 3,362 alliances in total, of which 1,205 alliances pertained to a clear R&D

component. Likewise, we focused merely on R&D acquisitions (Ahuja and Katila, 2001; Higgins

and Rodriguez, 2006), for which the business description in SDC Platinum suggested that the

acquired biotechnology company was indeed targeted for its R&D capabilities. Through this

process, 638 R&D acquisitions were found for the study period 1986–2000.

Control Variables

Presample Innovation In all models we attempted to capture the unobserved heterogeneity in

firm innovation. As stressed by Blundell et al. (1995), the ‗permanent‘ capabilities of companies

to successfully commercialize new products and processes should be reflected in the pre-sample

history of innovative output. Hence, we included both the pre-sample average patent count (i.e.,

the average number of biotech patents by the firm in the period from 197415

to 1985) and a

dummy (1 = if the firm had ever innovated prior to 1986)16

to proxy for the unobserved

heterogeneity in the innovation models of incumbent pharmaceutical firms (Blundell et al., 1995;

Blundell et al., 2002).

14 See Ahuja and Katila (2001), a simple count measure could possibly lead to measurement concerns. Instead of

considering the scale of external R&D activities, we focus on whether firms use external R&D or not. The average of

this dummy variable is 0.43. 15 Due to the emergence of the new biotechnology in the mid 1970s. 16 The dummy variable captures the fact that firms who sometimes innovate may be qualitatively different from those

who never innovate (Blundell et al., 1995).

14

Firm Merged The global pharmaceutical industry has witnessed a continuing trend of

consolidation and concentration, especially in the late 1990s. To account for horizontal mergers

and acquisitions among sample firms, as previously mentioned in the data section, we pursued a

detailed family tree analysis by tracing all incumbent pharmaceutical firms back to their

forerunners alive in 1986. Along with this analytical approach, we explicitly inserted a dummy

variable indicating whether or not the sample firm was the outcome of a horizontal merger or

acquisition over the study period 1986–2000 (1 = Firm Merged).

Pharmaceutical Firm The global pharmaceutical industry is made up of both specialized

pharmaceutical firms and more diversified, mainly chemical, conglomerates. A firm‘s level of

diversification indicates its previous experience in entering new businesses (Chang and Singh,

1999; Yip, 1982), which is likely to influence to what extent it attempts to innovate within the

new biotechnology regime. Therefore, we controlled for a firm‘s degree of diversification by

coding a firm as 1 if it is a specialized company (1 = Pharma Firm), and 0 otherwise. Specialized

pharmaceutical companies are those firms that are active in SIC 2834 (Pharmaceutical

Preparations Manufacturing), whereas a conglomerate might engage, for instance, in both SIC

2834 and SIC 2890 (Chemical Products Manufacturing). About one half of the firms are fully

specialized (45%).

Firm Nationality To control for country-specific institutional configurations that are

significant in shaping patenting propensities (Jaffe and Trajtenberg, 1999) as well as internal and

external R&D strategies, we included two indicator variables based upon the location of company

headquarters. One indicator variable is coded as 1 if the pharmaceutical firm is located in the

United States (1 = U.S. Firm); the other indicator is coded as 1 if the pharmaceutical firm is

headquartered in Europe (1 = European Firm), with a Japanese location as the reference category.

15

The global nature of our dataset is reflected in the fact that 36% of the firms are U.S. based, 28%

are European based, with the remaining 36% headquartered in Japan.

Firm Size Previous work has examined how factors associated with firm size impact on a

firm‘s incentive to invest in R&D (Cohen and Levinthal, 1989; Schumpeter, 1942; Teece, 1982,

1992). Moreover, firm size has been shown to exert a direct effect on innovative output (Acs and

Audretsch, 1988; Cohen and Klepper, 1996; Freeman and Soete, 1997). Researchers also observe

that small firms are more likely to restrict themselves to a simple innovation strategy, whereas

large firms tend to combine various R&D sources (Beneito, 2006; Cassiman and Veugelers,

1999). We thus controlled for firm size by taking into account the total number of employees

(Employees). The average number of employees of individual firms is 28,970, suggestive of their

large size.

Firm Performance Previous research suggests that there is a trade-off between the size of a

firm and its performance (Cubbin and Leech, 1986; Dobson and Gerrard, 1989; Reid, 1993,

1995). Companies with higher performance may as well be better positioned to finance their

R&D investments. These influences were captured by including Net Income and Total Revenues17

in the regression analysis to allow for financial status of large pharmaceutical companies.

Year Fixed Effects In order to control for factors other than firm-specific characteristics that

may influence all firms, such as economic-wide changes, we inserted year dummies, with 2000

being the reference year.

17 Total revenue is a measure of firm size as well and might account for firm size speedily increased through

horizontal mergers and acquisitions. Besides, control for firm revenue is necessary for isolating the effect of R&D

expenditures on innovative output (Rothaermel and Hess, 2007).

16

Estimation Technique

The dependent variable of this study, innovative output, as measured by biotech patent counts, is

a count variable taking only non-negative integer values. To model count data the linear

exponential or log-link family is a good alternative (Cameron and Trivedi, 1986). On account of

the overdispersion phenomenon displayed by patent counts, namely the preponderance of zero

and small values of firm patenting, the negative binomial estimation provides a better fit for the

data than the more restrictive Poisson model (Hausman et al., 1984).

The common first moment condition for the negative binomial model is,

(1)

In our innovation model, we write as

(2)

where λit is the expectation value of biotech patent counts for firm i in period t, RDI and RDI2 are

R&D expenditures and R&D expenditures squared, respectively, RDE represents the dummy

variable of external R&D through R&D alliances or R&D acquisitions, Controln is a vector of

control variables, and the parameters β, θ, γ are coefficients of various explanatory variables. A

one-period lag is employed on all explanatory (key and control) variables to alleviate a potential

simultaneity bias.

Differentiating equation (1) with respect to RDI and dividing both sides of the resulting

equation by λit yields,

(3)

17

The expected marginal innovative output is herein given by the relative change18

in λit associated

with a one unit change in RDI. We then differentiate equation (3) with respect to RDI a second

time,

(4)

The above model explains the concavity of marginal returns to internal R&D, indicating whether

there are decreasing returns (i.e., concave whenever or ) or increasing

returns (i.e., convex whenever or ) to scale in internal R&D. Further,

the existence of an interactive effect between RDI and RDE is defined by requiring that the

relative change in λit associated with a change in RDI depends on RDE and vice versa

(Winkelmann, 2008). Differencing equation (3) with respect to RDE, given a dummy variable of

RDE, we obtain19

,

(5)

Equation (5) clarifies the sense in which complementarity or substitutability between internal and

external R&D strategies is contingent on the level of in-house R&D. RDI and RDE are considered

(strict) complementary when the interactive effects are positive ( ), while

RDI and RDE are considered (strict) substitutive when the interactive effects are negative

( ).

18 See the interpretation of parameters in the Poisson regression by Winkelmann (2008, p.70). 19

Mullahy (1999) discusses the difficulties that arise if interactive effects in nonlinear models are defined in terms

of absolute, rather than relative, changes. In our model the interactive effects under absolute change would be

otherwise as complicated as follows,

18

In our count panel data model, either fixed or random effects specification can, in theory,

be used to control for firm-specific unobserved heterogeneity (Greene, 2003). One difficulty with

the random effects specification is its underlying assumption that time invariant unobserved

heterogeneity is uncorrelated with regressors of interest. In other words, this specification rules

out the existence of time invariant unobserved factors that influence both a firm‘s R&D strategies

and its innovative output, which is untenable since managers make their ‗strategic‘ organizational

choices with certain performance expectations in mind (Hamilton and Nickerson, 2003; Leiblein

et al., 2002). Hence, we used the fixed-effects specification so as to estimate the model

parameters consistently (Hamilton and Nickerson, 2003; Winkelmann, 2008). Additionally,

following Blundell et al. (1995, 2002), we allowed for time invariant unobserved heterogeneity

by explicitly taking into account the pre-sample history of firm innovative output.

RESULTS

Table 1 shows descriptive statistics and a bivariate correlation matrix. The table indicates a high

degree of variance on most of the variables such as Biotech Patents, R&D Expenditures,

Employees, and Presample Mean. The bivariate correlations are, with the exception of R&D

expenditures and employees20

, low and thereby indicate sound validity. Moreover, variance

inflation factors (VIFs) were computed to assess the severity of multicollinearity. The average

VIF value is 2.19, with the maximum VIF value of 5.29, which are well below the cut-off point

20 As mentioned previously, factors associated with firm size tend to influence a firm‘s incentive to invest in R&D

(Cohen and Levin, 1989; Schumpeter, 1942; Teece, 1982, 1992). In addition, employees and total revenue, as close

measures of firm size, are also highly correlated in our dataset (r = 0.80).

19

of 10 (Cohen et. al., 2003), implying that multicollinearity does not pose a problem for our

estimation models.21

Insert Table 1 about here

Table 2 provides results for all models using fixed-effects negative binomial estimation.

Model 1 is the baseline specification including the control variables only. Model 2 estimates

R&D expenditures and the quadratic terms thereof. Model 3 presents the full model with the

interactive effects between internal and external R&D. Each subsequent model demonstrates a

significant (p < 0.001) improvement over the baseline specification. We use the full Model 3 to

discuss the regression results of the hypothesis tests. Overall, the estimated coefficients suggest a

curvilinear relationship between R&D expenditures and biotech patenting. The estimated second

derivative, 2θ2 when RDE = 0 given by equation (4), is negative and statistically significant (p <

0.001). By comparison, when RDE = 1, the estimated second derivative, 2θ2 + 2θ23 given by

equation (4), becomes less negative (and significant at p < 0.10), implying that marginal returns

to internal R&D decrease less in the presence of external R&D. Taken together, these results

provide strong support for hypothesis 1 that there are decreasing returns to scale in internal R&D.

Firm innovative output initially increases with investment in internal R&D, this rate of increase

diminishes at higher levels of internal R&D spending though. This finding is consistent with

previous studies clearly suggesting diseconomies of scale in internal R&D (Graves and

21 As an additional test, we centered the independent variables before creating their squares and cross products

(Cohen, 2003). The estimated coefficients turned out to show a very similar empirical pattern. Besides, even if

multicollinearity does not bias coefficient estimates, it may affect the stability of the estimated coefficients, thereby

omitting even a few observations can change the sign or the significance of the affected variables (Greene, 2003). To

ensure the robustness of our results, we performed a sensitivity analysis by drawing random samples of 90% of the

total observations and estimating the full model for each of these random samples (Ahuja and Katila, 2001). The

corresponding results are similar to the results reported below.

20

Langowitz, 1993; Hagedoorn and Duysters, 2002; Lokshin et al., 2008; Rothaermel and Hess,

2007).

Insert Table 2 about here

In hypothesis 2, we postulated that internal and external R&D strategies may reinforce or

substitute for each other in predicting firms‘ innovative output. We tested the interactive effect

between internal and external R&D by examining the sign and statistical significance of the

values of the cross-partial derivative, given by equation (5). To this purpose, a

graphical analysis, as exhibited in Figure 1, was conducted to manifest the nature and

significance of the interactive effect by plotting its value and the implied z-statistic value over the

range of internal R&D investments by incumbent pharmaceutical firms (Ai and Norton, 2003;

Hoetker, 2007; Wiersema and Bowen, 2009). In Figure 1, the bullet symbols ( ● ) depict values of

the interactive effect between internal and external R&D, recorded on the left axis, while the

diamond shaped symbols ( ◊ ) depict z-statistic values, recorded on the right axis. As seen in

Figure 1, the sign, magnitude, and statistical significance of the interactive effect varies with the

level of in-house R&D. The interactive effect is negative and significant (p < 0.10) when internal

R&D is less than 886.14, whereas the effect is positive and significant (p < 0.10) when internal

R&D is in the range from about 1364.34 to 2689.31, with the interactive effect not statistically

significant in between. These findings highlight the complexity of understanding the relationship

between internal and external R&D strategies in shaping firms‘ innovative performance, and

clear-cut results, either complementarity (hypothesis 2-a) or substitutability (hypothesis 2-b), are

not always to be expected. In contrast to conventional wisdom, our overall findings suggest that a

21

complete picture is to be found in the contingency of the relationship between internal and

external R&D strategies, where complementarity or substitutability is contingent on the level of

in-house R&D.

Insert Figure1 about here

Turning to the effects of control variables, we found largely consistent results across

various models. The pre-sample patent variables, i.e., pre-sample mean patent count and the

dummy, are both positive and statistically significant, indicating that it is important to control for

the unobserved differences in the innovative capabilities with which firms entered our sample.

Employees, as a measure of firm size, are positively associated with patenting frequency (p <

0.05). The implied gains from large horizontal merger or acquisition, however, have not been

identified. Incumbent pharmaceutical companies that speedily expand through horizontal mergers

and acquisitions appear to degrade their innovative output significantly (p < 0.001) over the study

period. The estimated coefficient of firm revenues is significant (p < 0.001) but, contrary to

expectations, in a negative direction. With regard to the year fixed effects (included but not

shown), the year dummies for 1993 – 1999 are positive and significant, while all the other

calendar year indicators are not statistically significant, relative to the reference year 2000. These

results suggest that patenting activity has significantly accelerated in the later sample period

(1993 – 1999), in contrast with its earlier years (1986 – 1992).

22

Robustness of Results

To ensure the robustness of our results presented above, we also ran models including a measure

of firm knowledge stock22

, effectively a lagged dependent variable, to partially control for firm-

specific unobserved heterogeneity that is not fixed throughout time (Blundell et al., 1995;

Blundell et al., 2002; Dushnitsky and Lenox, 2005). Since the lagged dependent variable is

predetermined, the conditional maximum likelihood estimator for count panel data models is

inconsistent (Blundell et al., 1995; Blundell et al., 2002; Crépon and Duguet, 1997; Montalvo,

1997; Wooldridge, 1997). Following Blundell et al. (1995, 2002), we used the pre-sample mean

(PSM) estimator23

, which is a good alternative to standard estimators when there are

predetermined regressors. The results under PSM estimation are reported in the first and third

columns of Table 3. These estimated results set a very similar pattern as what we observed in

Table 2. The additional message is that pharmaceutical firms with higher knowledge stock

exhibit a significantly greater number of biotechnology patents (p < 0.001). The strong effects of

presample dummy and firm merged, though, are driven into insignificance in this dynamic

analysis.

Furthermore, as our study focuses on a dummy, rather than a count, variable of external

R&D, we disaggregate external R&D into R&D alliances and R&D acquisitions so as to capture

the scale of different types of external R&D investments. The corresponding results are shown in

the last two columns of Table 3. Both estimation methods — fixed-effects negative binomial and

PSM — produce fairly close findings to those presented before. The robust empirical patterns

overall suggest that the complementarity or substitutability between internal R&D and R&D

22 See Blundell et al. (1995), knowledge stock (Kit) is the depreciated sum of past innovations and defined as,

Kit = Yit + (1– δ) Kit-1 where the depreciation rate δ is taken to be 30%, Yit is biotech patent in our models. 23 Another alternative in the case of predetermined regressors is the quasi-differenced GMM estimator, which can

however be severely biased in small samples, particularly when regressors are highly persistent and the instruments

are therefore weak predictors of the endogenous variables in the differenced model (Blundell et al., 2002).

23

alliances, or between internal R&D and R&D acquisitions, is contingent on the level of in-house

R&D investments. The only noteworthy difference between these two methods of estimation,

fixed-effects negative binomial and PSM, is that R&D acquisitions exhibit more significant

interactive effects with internal R&D under PSM estimation. In sum, these results corroborate our

findings from the more parsimonious models of Table 2.

Insert Table 3 about here

DISCUSSION AND CONCLUSION

A substantial body of literature has examined the relationship between internal R&D and external

technology sourcing through external R&D, so far focusing on either complementarity or

substitutability. Yet there is still considerable confusion over the conditions under which there

may in fact be complementarity or substitutability between different R&D strategies. Following

Cassiman and Veugelers (2006), success in innovation will depend not only on combing various

innovation activities, but also on creating the right context.24

The potential benefits from

complementarity are therefore context-specific or contingent. One implication of our analysis is

that the level of in-house R&D investments, which is characterized by decreasing marginal

returns, is a contingency variable that critically influences the nature of the link between internal

and external R&D strategies. In particular, internal R&D and external R&D, through R&D

alliances or R&D acquisitions, are complementary innovation activities at higher levels of in-

house R&D investments, whereas at lower levels of in-house R&D efforts internal and external

R&D are substitutive strategic options.

24 In their study, Cassiman and Veugelers (2006) identify reliance on more basic R&D, i.e., the use of universities

and research centers as information sources for the innovation process, as an important context variable that affects

the strength of the complementarity between innovation activities.

24

In line with the two roles, or faces, of in-house R&D25

(Cohen and Levinthal, 1989, 1990),

a minimum level of absorptive capacity, derived from in-house R&D activities, is required for

effectively incorporating external R&D (Catozzella and Vivarelli, 2007; Lokshin et al., 2008;

Tsai and Wang, 2008). Higher levels of internal R&D investments, as a consequence, may enable

incumbent firms to maintain their ability to react adequately to technological changes by

leveraging external networks. As Arora and Gambardella (1990) demonstrate in the context of

pharma-biotechnology, large pharmaceutical firms with higher internal research activities are

more active in using external knowledge sources. In this light, sufficient investments in in-house

R&D contribute substantially to a spiral of success and thereby sustainable innovative

capabilities. By contrast, at lower levels of in-house R&D investments, our findings point to the

idea that the simultaneous pursuit of internal R&D and external R&D sourcing strategies would

actually decrease a firm‘s innovative output, at least at the margin. Part of the explanation may lie

in the fairly constrained resource set of firms, especially in terms of their absorptive capacity, as a

firm may fail to reap economies of scope when ineffectively scanning, selecting, and assimilating

external technological knowledge. Alternatively, incumbent firms with low research capabilities

tend to exhibit overreliance on external R&D alliances or acquisitions for promising technologies

in the face of radical technological change. This would explain, to a certain extent, why some

established pharmaceutical firms are obtaining or choose to obtain their most technologically

advanced projects from alliance partners or through the market for know-how.

In attempting to characterize the contingency of the relationship between internal and

external R&D strategies in shaping firms‘ innovative output, a number of econometric problems

arising from the count panel nature of the data have to be considered. In essence, we adopted an

25 As Cohen and Levinthal (1989, 1990) point out, the two roles, or faces, of internal R&D are reflected in directly

generating new knowledge and indirectly contributing to absorptive capacity.

25

estimation approach that places the simultaneous interactions between internal R&D and two

alternative modes of external R&D investments, R&D alliances and R&D acquisitions26

, at the

core of our analysis. It is readily apparent that firms do not randomly choose a specific way to

organize R&D, either rely exclusively on in-house R&D or systematically combine with R&D

alliances and/or R&D acquisitions. The effects that drive firms to choose ex ante a specific R&D

strategy towards superior performance outcomes27

, when not corrected for, would introduce a

serious endogeneity bias28

(Hamilton and Nickerson, 2003; Leiblein et al., 2002). Possible

remedies are available, nevertheless, if endogeneity is caused by time invariant correlated

unobserved heterogeneity, for which one can estimate the model parameters consistently by fixed

effects panel data methods (Winkelmann, 2008). Accordingly, we employed fixed effects

specification in our panel count data models. In addition, to partially control for firm-specific

unobserved heterogeneity that is not fixed throughout time (Blundell et al., 1995; Blundell et al.,

2002; Dushnitsky and Lenox, 2005), a dynamic analysis was performed by taking into account

the measure of firm knowledge stock.

So far the inconsistent literature on the complementarity or substitutability between R&D

strategies, as emphasized by Watkins and Paff (2009), leaves considerable room for improving

the consistency in defining and measuring complementarity, particularly at the boundaries

26 In some cases large pharmaceutical firms also hope to establish a ‗preferential link‘ with the new biotechnology

venture by acquiring part of its capital stocks. Such acquisition of minority stocks would likely translate into a formal

alliance agreement once a specific new discovery needs to be implemented and possibly commercialized (Arora and

Gambardella, 1990). 27 Understanding how firms choose ex ante a specific way to organize R&D is beyond the scope of our study. 28 Another concern is the sample selection bias. A varied set of sample selection models are available in econometric

literature to deal with such self-selection bias — the foremost model being Heckman two-stage approach. However,

while Heckman and similar approaches are most commonly used within the domain of continuous-response models,

they are not appropriate for count response models (Cameron and Trivedi, 1998; Hilbe, 2007). The development of

methodology in count panel data models to correct for sample selection bias is unfortunately still an open task

(Winkelmann, 2008).

26

between disciplinary traditions29

. The cloud of confusion over what is and what conditions

complementarity thus suggests further research work to disentangle the rationales for combining

different R&D sourcing arrangements. Moreover, there appears to be increasing variation among

the largest pharmaceutical firms in terms of their strategies to extract value from new

technologies, each trying to achieve a preferred balance between in-house R&D and externally

sourced R&D. It may consequently no longer be useful to talk about Big Pharma as a

homogenous sector (Mittra, 2007). An interesting avenue for future work would be to investigate

these issues, for instance, through a fine-grained analysis of firm‘s optimal balance between

internal and external R&D strategies at various stages of innovation process or towards different

types of innovation. We believe that a fuller appreciation of the origins of innovativeness is to be

found in a better understanding of the alignment between R&D strategies and the internal and

external environment of firms, be it in either a high-technology or a low-technology industry.

29

In the standard economic literature, partial substitution elasticity among inputs theoretically measures the relative

change in the ratio of inputs when their relative prices change. By contrast, a very different measure is recently used

in the management literature, comparing changes in average output ratios under different combinations of inputs. For

the latter measure input prices play no role, nor do margins (Watkins and Paff, 2009).

27

REFERENCES

Acs, Z.J., Audretsch, D.B. 1988. Innovation in large and small firms: An empirical analysis.

American Economic Review 78: 678-690.

Acs, Z.J., Audretsch, D.B. 1991. Innovation and size at the firm level. Southern Economic

Journal 57: 739–744.

Ahuja, G., Katila, R. 2001. Technological acquisitions and the innovation performance of

acquiring firms: A longitudinal study. Strategic Management Journal 22: 197-220.

Ai, C., Norton, E.C. 2003. Interaction terms in logit and probit models. Economics Letters 80:

123-129.

Arora, A., Gambardella, A., 1990. Complementarity and external linkages: The strategies of the

large firms in biotechnology. Journal of Industrial Economics 38: 361-379.

Audretsch, D.B., Menkveld, A.J., Thurik, A.R. 1996. The decision between internal and external

R&D. Journal of Institutional and Theoretical Economics 152: 519-530.

Beneito, P. 2006. The innovative performance of in-house and contracted R&D in terms of

patents and utility models. Research Policy 35: 502-517.

Bettis, R.A., Hitt, M.A. 1995. The new competitive landscape. Strategic Management Journal

(Special Issue) 16: 7-19.

Blonigen, B.A., Taylor, C.T. 2000. R&D intensity and acquisitions in high-technology industries:

Evidence from the US electronic and electrical equipment industries. Journal of Industrial

Economics 48: 47-70.

Blundell, R., Griffith, R., Reenen, J.V. 1995. Dynamic count data models of technological

innovation. Economic Journal 105: 333-344.

Blundell, R., Griffith, R., Windmeijer, F. 2002. Individual effects and dynamics in count data

models. Journal of Econometrics 108: 113-131.

Breschi, S., Malerba, F. 1997. Sectoral systems of innovation: Technological regimes,

Schumpeterian dynamic, and spatial boundaries. In: Edquist, C. (ed.). Systems of

Innovation. London: Pinter.

28

Brown, S.L., Eisenhardt, K.M. 1997. The art of continuous change: Linking complexity theory

and time-paced evolution in relentlessly shifting organizations. Administrative Science

Quarterly 42: 1-34.

Caloghirou, Y., Kastelli, I., Tsakanikas, A. 2004. Internal capabilities and external knowledge

sources: Complements or substitutes for innovative performance? Technovation 24: 29-39.

Cameron, A.C., Trivedi, P.K. 1986. Econometric models based on count data: Comparisons and

applications of some estimators and tests. Journal of Applied Econometrics 1: 29-54.

Cappelen, A., Raknerud, A., Rybalka, M. 2008. The effects of R&D tax credits on patenting and

innovation. Statistics Norway, Research Department Discussion Paper No. 565.

Cassiman, B., Veugelers, R. 1999. Make and buy in innovation strategies: Evidence from

Belgian manufacturing firms. Research Policy 28: 63-80.

Cassiman, B., Veugelers, R. 2006. In search of complementarity in innovation strategy: Internal

R&D and external technology acquisition. Management Science 52: 68-82.

Catozzella, A., Vivarelli, M. 2007. The catalyzing role of in-house R&D in fostering the

complementarity of innovative inputs. IZA Discussion Paper No. 3126.

Chang, S.J., Singh, H. 1999. The impact of modes of entry and resource fit on modes of exit by

multibusiness firms. Strategic Management Journal 20: 1019–1035.

Chesbrough, H.W. 2003. Open Innovation: The New Imperative for Creating and Profiting from

Technology. Boston: Harvard Business School Press

Christensen, C.M. 2000. The Innovator’s Dilemma. New York: HarperBusiness.

Cohen, P., Cohen, J., West, S.G., Aiken, L.S. 2003. Applied Multiple Regression/Correlation

Analysis for the Behavioral Science. 3rd

ed. Hillsdale, NJ: Erlbaum.

Cohen, W.M., Klepper, S. 1996. A reprise of size and R&D. Economic Journal 106: 925-951.

Cohen, W.M., Levinthal, D.A. 1989. Innovation and learning: The two faces of R&D. Economic

Journal 99: 569-596.

Cohen, W.M., Levinthal, D.A. 1990. Absorptive capacity: A new perspective on learning and

innovation. Administrative Science Quarterly 35: 128-152.

29

Crépon, B., Duguet, E. 1997. Estimating the innovation function from patent numbers: GMM on

count panel data. Journal of Applied Econometrics 12: 243-264.

Cubbin, J., Leech, D. 1986. Growth versus profit maximisation: A simultaneous-equations

approach to testing the Marris model. Managerial and Decision Economics 7: 123-131.

Czarnitzki, D., Hanel, P., Rosa, J. 2004. Evaluating the impact of R&D tax credit on innovation:

A microeconometric study on Canadian firms. ZEW discussion paper 04-77.

Darby, M.R., Zucker, L.G. 2003. Grilichesian breakthroughs: Inventions of methods of inventing

and firm entry in nanotechnology. Cambridge, MA. NBER Working Paper No. 9825.

Dobson, S.M., Gerrard, B.J. 1989. Growth and profitability in the Leeds engineering sector.

Scottish Journal of Political Economy 36: 334-352.

Dushnitsky, G., Lenox, M.J. 2005. When do incumbents learn from entrepreneurial ventures?

Corporate venture capital and investing firm innovation rates. Research Policy 34: 615-

639.

Duysters, G., Hagedoorn, J. 1993. The Cooperative Agreements and Technology Indicators

(CATI) Information System. Maastricht: MERIT.

Duysters, G., Hagedoorn, J. 2001. Do company strategies and structures converge in global

markets? Journal of International Business Studies 32: 347-356.

Edquist, C. 1997. Systems of Innovation: Technologies, Institutions, and Organizations. London:

Printer.

Foster, R.N., Kaplan, S. 2001. Creative Destruction. New York: Doubble-day.

Freeman, C., Soete, L. 1997. The Economics of Industrial Innovation. London: Printer.

Graves, S.B., Langowitz, N.S. 1993. Innovative productivity and returns to scale in the

pharmaceutical industry. Strategic Management Journal 14: 593-605.

Greene, W.H. 2003. Econometric Analysis. 4th

ed. New York: Prentice Hall.

Griffith, R., Redding, S., van Reenen, J. 2004. Mapping the two faces of R&D: Productivity

growth in a panel of OECD countries. Review of Economics and Statistics 86: 883-895.

Griliches, Z. 1979. Issues in assessing the contribution of research and development to

productivity growth. Bell Journal of Economics 10: 92-116.

30

Griliches, Z. 1990. Patent statistics as economic indicators: A survey. Journal of Economic

Literature 28: 1661-1707.

Griliches, Z. 1998. R&D and Productivity – The Econometric Evidence. Chicago: University of

Chicago Press.

Guo, J.Q., Trivedi, P.K. 2002. Flexible parametric models for long-tailed patent count

distributions. Oxford Bulletin of Economics and Statistics 63: 63-82.

Gurmu, S., Perez-Sebastian, F. 2007. Patents, R&D, and lag effects: Evidence from flexible

methods for count panel data on manufacturing firms. Andrew Young School of Policy

Studies Working Paper Series.

Hagedoorn, J. 1993. Understanding the rationale of strategic technological partnering:

Interorganizational modes of cooperation and sectoral differences. Strategic Management

Journal 14: 371-385.

Hagedoorn, J., Cloodt, M. 2003. Measuring innovative performance: Is there an advantage in

using multiple indicators? Research Policy 32: 1365-1379.

Hagedoorn, J., Duysters, G. 2002. The effect of mergers and acquisitions on the technological

performance of companies in a high-tech environment. Technology Analysis & Strategic

Management 14: 68-85.

Hall, A.J., Sivamohan, M.V.K., Clark, N., Taylor, S., Bockett, G. 2001. Why research

partnerships really matter: Innovation theory, institutional arrangements and implications

for the developing new technology for the poor. World Development 29: 783–797.

Hall, B.H., Griliches, Z., Hausman, J. 1986. Patents and R&D: Is there a lag? International

Economic Review 27: 265-283.

Hamiton, B.H., Nickerson, J.A. 2003. Correcting for endogeneity in strategic management

research. Strategic Organization 1: 51-78.

Hausman, J., Hall, B.H., Griliches, Z. 1984. Econometric models for count data with an

application to the patents-R&D relationship. Econometrica 52: 909-938.

Henderson, R., Cockburn, I. 1994. Measuring competence? Exploring firm effects in

pharmaceutical research. Strategic Management Journal 15: 63-84.

31

Higgins, M., Rodriguez, D. 2006. The outsourcing of R&D through acquisition in the

pharmaceutical industry. Journal of Financial Economics 80: 351-383.

Hilbe, J.M. 2007. Negative Binomial Regression. Cambridge: Cambridge University Press.

Hitt, M.A., Hoskisson, R.E., Kim, H. 1997. International diversification: Effects of innovation

and firm performance in product-diversified firms. Academy of Management Journal 40:

767– 798.

Hoetker, G. 2007. The use of probit and logit models in strategic management research: Critical

issues. Strategic Management Journal 28: 331-343.

Jaffe, A.B., Trajtenberg, M. 1999. International knowledge flows: Evidence from patent citations.

Economics of Innovation & New Technology 8: 105–136.

Knott, A.M., Posen, H.E. 2009. Firm R&D behavior and evolving technology in established

industries. Organization Science 20: 352-367.

Kranenburg, H.L., Hagedoorn, J. 2008. Strategic focus of incumbents in the European

telecommunications industry: The cases of BT, Deutsche Telekom and KPN.

Telecommunications Policy 32: 116-130.

Laursen, K., Salter, A. 2006. Open for innovation: The role of openness in explaining innovation

performance among U.K. manufacturing firms. Strategic Management Journal 27: 131-

150.

Leiblein, M.J., Reuer, J.J., Daisace, F. 2002. Do make or buy decisions matter? The influence of

organizational governance on technological performance. Strategic Management Journal

23: 817-833.

Lokshin, B, Belderbos, R., Carree, M. 2008. The productivity effects of internal and external

R&D: Evidence from a dynamic panel data model. Oxford Bulletin of Economics and

Statistics 70: 399-413.

Love, J.H., Roper, S. 2002. Internal versus external R&D: A study of R&D choice with sample

selection. International Journal of the Economics of Business 9: 239-255.

Milgrom, P., Roberts, J. 1990. The economics of modern manufacturing: Technology, strategy,

and organization. American Economic Review 80: 511-528.

32

Milgrom, P., Roberts, J. 1995. Complementarities and fit: Strategy, structure, and organizational

change in manufacturing. Journal of Accounting and Economics 19: 179-208.

Mittra, J. 2007. Life science innovation and the restructuring of the pharmaceutical industry:

Merger, acquisition and strategic alliance behavior of large firms. Technology Analysis

and Strategic Management 19: 279-301.

Mohnen, P., Röller, L. 2005. Complementarities in innovation policy. European Economic

Review 49: 1431-1450.

Montalvo, J.G. 1997. GMM estimation of count-panel-data models with fixed effects and

predetermined instruments. Journal of Business & Economic Statistics 15: 82-89.

Mullahy, J. 1999. Interaction effects and difference-in-difference estimation in loglinear models.

NBER Technical Working Paper No. 245.

Nagarajan, A., Mitchell, W. 1998. Evolutionary diffusion: Internal and external methods used to

acquire encompassing, complementary, incremental technological changes in the

lithotripsy industry. Strategic Management Journal 19: 1063-1077.

Nicholls-Nixon, C.L., Woo, C.Y. 2003. Technology sourcing and output of established firms in a

regime of encompassing technological change. Strategic Management Journal 24: 651-

666.

Parsons, M., Phillips, N. 2007. An evaluation of the federal tax credit for scientific research and

experimental development. Department of Finance, Canada, Working Paper 2007-08.

Pisano, G.P. 1997. The Development Factory: Unlocking the Potential of Process Innovation.

Boston, MA: Harvard Business School Press.

Powell, W.W. 1998. Learning from collaboration: Knowledge and networks in the biotechnology

and pharmaceutical industries. California Management Review 40: 228-240.

Powell, W.W., Koput, K.W., Smith-Doerr, L. 1996. Interorganizational collaboration and the

locus of innovation: Networks of learning in biotechnology. Administrative Science

Quarterly 41: 116-145.

Reid, G.C. 1993. Small Business Enterprise: An Economic Analysis. London: Routledge.

33

Reid, G.C. 1995. Early life-cycle behavior of micro-firms in Scotland. Small Business Economics

7: 89-95.

Rothaermel, F.T., Hess, A.M. 2007. Building dynamic capabilities: Innovation driven by

individual-, firm-, and network-level effects. Organization Science 18: 898-921.

Scherer, F.M. 1982. Inter-technology flows and productivity growth. Review of Economics and

Statistics 64: 627-634.

Schmiedeberg, C. 2008. Complementarities of innovation activities: An empirical analysis of the

German manufacturing sector. Research Policy 37: 1492-1503.

Schumpeter, J.A. 1942. Capitalism, Socialism and Democracy. New York: Harper.

Stuart, T.E. 2000. Interorganizational alliances and the performance of firms: A study of growth

and innovation rates in a high-technology industry. Strategic Management Journal 21:

791-811.

Teece, D. J. 1982. Towards an economic theory of the multiproduct firm. Journal of Economic

Behavior and Organization 3: 39-63.

Teece, D. J. 1988. Technological change and the nature of the firm. In: Dosi, G., Freeman, C.,

Nelson, R., Silverberg, G., Soete, L. (ed.). Technical Change and Economic Theory.

London: Pinter Publisher.

Teece, D.J. 1992. Competition, cooperation, and innovation: Organizational arrangements for

regimes of rapid technological progress. Journal of Economic Behavior and Organization

18: 1-25.

Trajtenberg, M. 1990. A penny for your quotes: Patent citations and the value of innovations.

RAND Journal of Economics 20: 172-187.

Tripsas, M. 1997. Unraveling the process of creative destruction: Complementary assets and

incumbent survival in the typesetter industry. Strategic Management Journal (Special

Issue) 18: 119-142.

Tsai, K., Wang, J. 2008. External technology acquisition and firm performance: A longitudinal

study. Journal of Business Venturing 23: 91-112.

34

Veugelers, R. 1997. Internal R&D expenditures and external technology sourcing. Research

Policy 26: 303-315.

Walker, R.D. 1995. Patents as Scientific and Technical Literature. Metuchen, NJ, and London:

The Scarecrow Press.

Watkins, T.A., Paff, L.A. 2009. Absorptive capacity and R&D tax policy: Are in-house R&D and

external contract R&D substitutes or complements? Small Business Economics 33: 207-

227.

Wiersema, M.F., Bowen, H.P. 2009. The use of limited dependent variable techniques in strategy

research: Issues and methods. Strategic Management Journal 30: 679-692.

Winkelmann, R. 2008.Econometric Analysis of Count Data. 5th

ed. Berlin: Springer-Verlag.

Wooldridge, J.M. 1997. Multiplicative panel data models without the strict exogeneity

assumption. Econometric Theory 13: 667-679.

Yip, G. 1982. Diversification entry: Internal development versus acquisition. Strategic

Management Journal 3: 331-345.

Table 1 Descriptive Statistics and Bivariate Correlation Matrix

Variable

Mean S. D. Min Max 1 2 3 4 5 6 7 8 9 10 11 12

1

Biotech Patents 7.60 14.42 0.00 207.00 1.00

2

R&D Expenditures (MM$) t-1* 492.17 557.26 0.30 2869.31 0.38 1.00

3

R&D External t-1 0.43 0.50 0.00 1.00 0.27 0.33 1.00

4

Firm Merged t-1 0.07 0.26 0.00 1.00 0.31 0.34 0.17 1.00

5

European Firm t-1 0.28 0.45 0.00 1.00 0.09 0.21 0.01 0.14 1.00

6

US Firm t-1 0.36 0.48 0.00 1.00 0.09 0.19 0.14 -0.02 -0.46 1.00

7

Pharma Firm t-1 0.45 0.50 0.00 1.00 0.19 -0.06 0.21 0.22 -0.08 -0.06 1.00

8

Employees (1,000) t-1 28.97 32.99 0.32 173.00 0.18 0.77 0.14 0.15 0.28 0.25 -0.31 1.00

9

Net Income (MM$) t-1* 595.69 963.44 -6680.33 7948.49 0.28 0.69 0.23 0.16 0.05 0.33 -0.09 0.57 1.00

10

Revenues (MM$) t-1* 9377.54 10729.70 15.83 53885.29 0.07 0.61 0.03 0.03 0.21 0.21 -0.42 0.80 0.57 1.00

11

Presample Meant-1 3.64 5.73 0.00 27.38 0.56 0.40 0.29 0.45 -0.08 0.21 0.28 0.15 0.34 0.05 1.00

12

Presample Dummyt-1 0.91 0.29 0.00 1.00 0.04 -0.17 -0.12 -0.07 -0.28 0.09 -0.04 -0.13 -0.06 -0.01 0.20 1.00

Notes: N = 1,139 firm-year observations. *In constant, inflation-adjusted year 2000 US $.

Table 2 Fixed-Effects Negative Binomial Regression Predicting Biotech Patenting

Variable Model 1 Model 2 Model 3

Constant 0.1134

(0.3374)

-0.0927

(0.3473)

-0.2953

(0.3545)

Year Fixed Effects

Included Included Included

Firm Merged t-1 -0.9764***

(0.2746)

-1.1353***

(0.2669)

-1.1584***

(0.2650)

European Firm t-1 -1.3036***

(0.2726)

-1.2720***

(0.2803)

-1.2603***

(0.2828)

US Firm t-1 -0.8015**

(0.2764)

-0.9207**

(0.2854)

-0.9156**

(0.2865)

Pharma Firm t-1 0.0839

(0.2270)

-0.0305

(0.2232)

0.0153

(0.2236)

Employees t-1 0.0166***

(0.0034)

0.0076*

(0.0039)

0.0079*

(0.0038)

Net Income t-1 1.49E-05

(4.05E-05)

-1.32E-05

(3.90E-05)

-1.40E-05

(3.93E-05)

Revenues t-1 -3.91E-05**

(1.25E-05)

-4.77E-05***

(1.34E-05)

-4.93E-05***

(1.32E-05)

Presample Meant-1 0.0528**

(0.0161)

0.0291†

(0.0171)

0.0321†

(0.0176)

Presample Dummy t-1 1.3161***

(0.2733)

1.3288***

(0.2920)

1.3900***

(0.2911)

R&D Expenditures t-1 1.70E-03***

(3.33E-04)

2.30E-03***

(3.81E-04)

R&D Expenditures squared t-1 -4.07E-07***

(1.14E-07)

-6.99E-07***

(1.49E-07)

R&D External t-1 0.2555*

(0.0993)

R&D Expenditures*R&D External t-1 -9.36E-04**

(2.82E-04)

R&D Expenditures Squared*

R&D External t-1

4.36E-07**

(1.41E-07)

N 1125 1125 1125

Log Likelihood -2273.8789 -2258.3008 -2252.8549

Chi Square 286.36*** 337.51*** 351.57***

Notes: Standard errors are in parentheses; year dummies are included but not shown. The fixed-effects

specification eliminates firms who never patent, thereby reducing the effective sample to 1125 firm-year

observations. Two tailed-tests significant at:

† p<0.10

* p<0.05

** p<0.01

*** p<0.001

37

Figure 1 Interactive Effects between Internal and External R&D

Under Relative Change

-4-2

02

4

z_statistic

-.001

-.0005

0

.0005

.001

.0015

Interaction

0 1000 2000 3000RD_Internal

Interaction z_statistic

Notes: z-statistic value (-1.645, 1.645), 10% two-tailed test. The solid symbols depict values of the

interaction effect between internal and external R&D, recorded on the left axis, while the diamond shaped

symbols depict z-statistic values, recorded on the right axis. The values of the interaction effect, as given by

equation (5), range from -0.0009 to 0.0016, with a mean value of -0.0005; the z-statistic values range from

-3.3142 to 2.8260, with a mean value of -2.3453.

38

Table 3 Robustness Test

Variable

External R&D (dummy) External R&D (count)

PSM Negative Binomial (FE) PSM

Constant

0.2136

(0.4545)

-0.1925

(0.3504)

0.1899

(0.4184)

Year Fixed Effects

Included Included Included

Firm Merged t-1

-0.0527

(0.1908)

-1.1806***

(0.2712)

-0.0431

(0.1693)

European Firm t-1

0.0996

(0.247)

-1.2490***

(0.2752)

0.0370

(0.2322)

US Firm t-1

-0.1470

(0.2339)

-0.9321**

(0.2807)

-0.2045

(0.2249)

Pharma Firm t-1

-0.0883

(0.1932)

-0.0282

(0.2258)

-0.1152

(0.1870)

Employees t-1

0.0086*

(0.0040)

0.0100*

(0.0040)

0.0087*

(0.0042)

Net Income t-1

-3.20E-06

(6.82E-05)

-8.15E-06

(3.88E-05)

3.73E-05

(5.84E-05)

Revenues t-1

-3.61E-05**

(1.28E-05)

-5.08E-05***

(1.33E-05)

-3.66E-05**

(1.32E-05)

Presample Mean t-1

0.0393**

(0.0137)

0.0394*

(0.0178)

0.0469**

(0.0147)

Presample Dummy t-1

-0.0623

(0.2749)

1.2789***

(0.2935)

0.0276

(0.2806)

Knowledge Stockt-1

0.0098***

(0.0017)

0.0094***

(0.0017)

R&D Expenditures t-1

2.12E-03***

(6.31E-04)

1.93E-03***

(3.58E-04)

1.97E-03***

(5.32E-04)

R&D Expenditures squared t-1

-9.52E-07***

(2.56E-07)

-5.57E-07***

(1.25E-07)

-8.60E-07***

(2.07E-07)

R&D External t-1

0.6121**

(0.2191)

R&D Expenditures*

R&D External t-1

-1.22E-03*

(5.64E-04)

R&D Expenditures Squared*

R&D External t-1

6.28E-07*

(2.67E-07)

R&D Alliancet-1

0.1120**

(0.0390)

0.1964***

(0.0561)

R&D Expenditures*

R&D Alliancet-1

-2.05E-04**

(6.15E-05)

-3.02E-04*

(1.17E-04)

R&D Expenditures Squared*

R&D Alliancet-1

9.11E-08***

(2.35E-08)

1.27E-07**

(4.68E-08)

R&D Acquisitiont-1

0.1076

(0.0736)

0.3139**

(0.0995)

R&D Expenditures*

R&D Acquisition t-1

-2.04E-04†

(1.19E-04) -5.56E-04***

(1.64E-04)

R&D Expenditures Squared*

R&D Acquisitiont-1

7.88E-08†

(4.74E-08)

2.03E-07**

(6.73E-08)

N 1139 1125 1139

Log Likelihood -2247.5532

Chi Square 360.39***

Notes: Standard errors are in parentheses; year dummies are included but not shown. The fixed-effects (FE)

specification eliminates firms who never patent, thereby reducing the effective sample to 1125 firm-year observations.

Two tailed-tests significant at: † p<0.10, * p<0.05, ** p<0.01, *** p<0.001.

The UNU-MERIT WORKING Paper Series 2010-01 Endogenous Economic Growth through Connectivity by Adriaan van Zon and Evans

Mupela 2010-02 Human resource management and learning for innovation: pharmaceuticals in Mexico

by Fernando Santiago 2010-03 Understanding multilevel interactions in economic development by Micheline Goedhuys

and Martin Srholec 2010-04 The Impact of the Credit Crisis on Poor Developing Countries and the Role of China in

Pulling and Crowding Us Out by Thomas H.W. Ziesemer 2010-05 Is there complementarity or substitutability between internal and external R&D

strategies? by John Hagedoorn and Ning Wang